MMEB VisDoc: leaderboard
MMEB VisDoc — visual-document retrieval split of the Massive Multimodal Embedding Benchmark; embedding quality on screenshots, PDFs and rendered documents.
Metric: Visual Document Score (%). Source: huggingface.co. 101 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Ops-Colqwen3-4B | 84.12 | |
| 2 | DME-Large | 83.36 | |
| 3 | WeMM-Embedding-9B | 83.34 | |
| 4 | Taichu-UniRetriever-8B | 83.25 | |
| 5 | Ovis-VL-Embedding-9B | 83.06 | |
| 6 | seed1.6-embedding-1215 | 82.38 | |
| 7 | Qwen3-VL-Embedding-8B | 82.36 | |
| 8 | WeMM-Embedding-4B | 82.04 | |
| 9 | DME-Medium | 82.02 | |
| 10 | Wianmu-5B-v0 | 81.99 |
Interactive version: theaggregate.ai/benchmark?slug=mmeb-visdoc · How It Works · Data refreshed daily, snapshot 2026-10-20.